Radivilova T. Stochastic models and methods for analysis of self-similar network traffic

The method of finding the unbiased estimator for the Hurst exponent is offered in work using the technique of rescaled range and change of variance for an aggregated series. Load of the channel is simulated and the origin of queues in the buffer of a network traffic realization is investigated. Mathematical model of a self-similar traffic is founded and examined. It has been described as a self-similar random process that appears to be some functional transformation of fractal Gaussian noise. The suggested method of estimation for a network channel work-load dependence of the traffic parameters allows us to determine the maximum possible loading of this network at given value of buffer memory and flow capacity of the channel. The results of this imitation design give the base to development of prognostication methods for overcoming the overload in data communication systems.